Rejuvenating Memory Neurons Restores Learning in Aged and Alzheimer’s Mice

Picture an old mouse, dropped into a pool of milky water, searching for a hidden platform it learned about days ago. It swims in aimless loops, circling the edges, unable to recall where safety is. The memory was formed; somewhere in its brain, the neurons that encoded that experience still exist. They just can’t do their job anymore.

Now imagine giving those specific neurons a brief genetic tune-up. Not replacing them, not stimulating them with light, but nudging them back towards a younger molecular state. That’s essentially what a team at the Ecole Polytechnique Federale de Lausanne (EPFL) in Switzerland has done, and the results are striking.

Gabriel Berdugo-Vega, Johannes Graff and colleagues at EPFL’s Brain Mind Institute targeted what neuroscientists call engram cells – the sparse clusters of neurons that fire together during learning and get reactivated when we recall a memory. These cells are, in the language of the field, the brain’s memory trace. In aged mice and mouse models of Alzheimer’s disease, engrams tend to malfunction; they don’t reactivate properly during recall, and memory suffers as a consequence. The team asked a deceptively simple question: what if you could make those particular neurons young again?

Their approach draws on partial cellular reprogramming, a technique that’s been gaining serious traction in regenerative medicine over the past few years. It uses three genes — Oct4, Sox2 and Klf4, collectively known as OSK — that can wind back some of the molecular hallmarks of aging in cells without erasing their identity entirely.

Previous work had shown OSK could rejuvenate tissues elsewhere in the body, most notably restoring vision in a mouse model of blindness. And a handful of studies had found that broadly expressing reprogramming factors from a young age could help preserve cognitive abilities during aging. But nobody had tried something this precise: targeting OSK specifically to the neurons that encode a given memory, and doing it after cognitive decline had already set in.

The team used gene therapy vectors delivered by precise brain injections, combining a system that tags neurons active during learning with a genetic switch that briefly turns on OSK during a defined window. It’s clever engineering – the reprogramming factors only get expressed in the cells that fire when the animal is actually learning, and only for a short burst.

In aged mice tested on a fear conditioning task, animals that received OSK in their hippocampal engram cells recovered memory performance to levels seen in young controls. The reprogrammed neurons were preferentially reactivated during recall, and showed molecular signs of a younger state, including restored heterochromatin markers and improved nuclear architecture. Crucially, the cells kept their neuronal identity; they didn’t de-differentiate into something else. If anything, identity markers got stronger. When the team targeted engrams in the prefrontal cortex instead, remote memories formed weeks earlier were also rescued. Same approach, different brain region, similar outcome.

Things got more interesting with Alzheimer’s models. Using APP/PS1 mice in a water maze task, Berdugo-Vega and colleagues found that reprogramming hippocampal engrams improved the animals’ learning strategies during training – they shifted from random, non-spatial swimming to the efficient, hippocampally driven navigation typical of healthy mice. Targeting prefrontal cortex engrams restored long-term spatial memory.

Digging into the molecular details with single-nucleus sequencing, the team found that Alzheimer’s engram cells showed broad transcriptional disruptions affecting synaptic function, immune responses and cell identity. OSK treatment partly reversed these patterns. One particularly telling finding involved a potassium channel gene called Kcnj3. Its expression was suppressed in Alzheimer’s engrams, and electrophysiology recordings showed those neurons were hyperexcitable – firing too readily, a known problem in Alzheimer’s. After reprogramming, Kcnj3 expression bounced back and neuronal excitability normalized.

Perhaps the most intriguing part of the study, though, is what the team calls a cognitive clock. They built a regression model that predicts a mouse’s age from its spatial learning behaviour alone: how it navigates, how efficiently it searches, how precisely it homes in on a target. Alzheimer’s mice registered as cognitively older than their actual age on this clock. But after engram reprogramming, their predicted age dropped back to match their chronological age. In an unbiased clustering analysis of 105 mice across all experiments, OSK-treated animals consistently shifted towards the learning patterns of young, healthy mice.

It is, to be clear, a proof of concept. The team used only amyloid-based Alzheimer’s models (not tau models), examined a single time point, and the longest they tracked benefits was two weeks. How durable the effects might be, whether reprogrammed engrams influence neighbouring circuits, and how any of this could eventually translate to humans are all open questions. Still, the implication is worth sitting with: in aged and diseased brains, the engram cells that hold our memories aren’t broken beyond repair. They’re constrained. And with the right molecular nudge, at least in mice, they can be coaxed back to work.

Study link: https://www.cell.com/neuron/fulltext/S0896-6273(25)00925-0